In this situation, we will use PIO main algorithm to establish an optimal set of mutations allowing to reach an optimal informativity in the preloaded AML datasets.
PIO main algorithm details
PIO main algorithm goes through different steps:
1. Preliminary step: Select the mutation with the highest metric in the overall cohort as the first mutation.
2. Current step/mutation statistics Calculate step_UP and step_UPKB, corresponding to the number of new unique patients (or new unique patient per kilobase) that are detected with the selected mutation. This step is performed on the remaining cohort (i.e. patients already presenting a mutation with the current panel are removed).
3. Next Mutation selection For this step, PIO runs two optimization module in parallel: –> Informativity optimization: Select the mutation with the maximum selected metric in the remaining cohort (i.e. removing patients already presenting a mutation with the current panel)
–> Comutation optimization: Select the mutation with the maximum selected metric among patients with at least 1 mutation but less than n mutations with the current panel in the overall cohort. (n corresponding to the min-mutation/patient parameter).
The mutation with the highest metric (either in term of new patients diagnosed or additional patients with comutation) is selected and added to the panel.
4. Reiterate 2 & 3: The algorithm reiterates until no more patient are added with both step 3 approaches.
Analysis parameters & run
In this case, we will start an analysis with the following parameters:
- Preloaded datasets: AML
- Analysis mode: PIO optimal
- Group mutations by: Exon/intron
- Informativity metric: UPKB
- Min. patients/mutations: 2
- Min. mutations/patient: 3
Clicking on “run analysis” will lead you to the following results:
Overall graph
The main grah shows the overall performances of the selected panel in the overall cohort (merged cohorts from the preloaded AML datasets in this case). The total number of patients (y-axis) with at least 1,2,3,4 or 5 mutations is represented according to the total panel size (in kb) on the x-axis. Graphical parameters can be tweaked using the box on the right of the graph.

The proposed panel of ~350kb allows to catch 79% of AML patients with at least 3 mutations, and 98% with at least 1 mutation. Using the graphical parameters, we can easily restrict the panel to 200kb, reaching 77% of patients with at least 3 mutations.
Main panel
Below the overall graph, you will find a downloadable table with the proposed panel and several parameters. This table can be easily downloaded and filtered to reach the desired length or explore the proposed panel.
- UP and UPKB correspond to the respective informativity metric computed on the overall cohort.
- step_UP and step_UPKB correspond to the respective informativity metrics computed at the corresponding step (i.e. on the remaining cohort).
- step_UP_comut and step_UPKB_comut correspond to the number of unique patients or unique patients per kilobase added at the corresponding step among patients with at least 1 mutation but less than n mutations in the step panel.
- n_comut_n and p_comut_n correspond to the number (or percentage) of patients with at least n mutations on the overall cohort.
- cum_length corresponds to the cumulative length of the panel.
Individual cohorts
In this tab, the same analysis is performed individually for each cohort. The proposed panel can also be downloaded as a table for all cohorts, and then filtered using the “cohort” column.
Mutation exploration
Several additional statistics can be found in the “mutation” module.
- The “statistics” and “type & frequencies” tabs output mutations statistics computed on the overall cohort, without considering the optimal panel.
- The “Distribution” tab shows mutation co-occurence among genes or exons retained in the panel.